← Back to home

⟩ Ship agents fast — and in control.

Idea to production in weeks, not quarters — the right user, a budget you set, an off switch, and an answer to "who did what." All of it costs less than one runaway agent's bill.

Start free — individual developer

Individual Developer

$0 bring your own keys, local-first

Everything you need to build, run, and understand agents — on your machine, with your own provider keys.

  • Multi-engine IDE — Claude Code, Codex, LangChain, or your own
  • agenthippo CLI — serve, automate, CI
  • Any model — bring your own keys, or run local and self-hosted models
  • Spotlight analytics — traces, cost, charts, local-first
  • Agent Anywhere — Slack, WhatsApp, Telegram, APIs
  • Agent Packs — build, version, share
Download AgentHippo

Prefer hosted models with no keys? Sign in for welcome credits and top up as you go at credits.agenthippo.ai.

Services

Production Sprint

Three weeks. One business-impact agent. Measured and under control.

We start inside your business: map the workflows, pick the one where an agent moves a real number, and baseline it — cycle time, manual touches, cost per completed task. Then we deploy the agent in your environment and measure the same workflow again. Before handoff, our control checklist passes live.

  • Before/after metrics: time, touches, errors, cost
  • Agent acts under each user's real permissions
  • Hard spend limit — proven refused at the cap
  • Kill switch and version rollback, demonstrated live
  • Complete action record, exported in minutes
  • Runs in your environment; no shared superuser credential

Pricing: Fixed fee — typically $25,000, scoped with you up front — and fully creditable toward your first year of AgentHippo. Taking a few design partners this quarter on founding terms, in exchange for a case study — ask.

How the three weeks run

For production teams

Governed Deployment

from $500 per agent / month

Your agent, your infrastructure, our control plane. The agent runs on your VM or cloud with no provider key on the host; governance is managed for you.

  • Leases, kill switch, deny rules, and caps — hosted control plane
  • Hard per-agent spend budgets at the model gateway
  • Per-user cost attribution on every call
  • Signed agent versions with one-command rollback
  • Evidence trail shipped and queryable
  • On-call alerts — fleet health to your Grafana IRM or PagerDuty
  • Multiple agents, SSO — all your users included, no seat fees

Full perimeter

Team Perimeter

from $3,000 per month

The whole governed stack on your VM: identity edge, model gateway, control plane, and evidence archive — deployed and verified by scripts, supported by us.

  • Everything in Governed Deployment
  • Per-user data access — Postgres RLS, DynamoDB, Databricks Unity Catalog, and more
  • Agents never hold a database or provider credential
  • WORM evidence archive in your storage
  • Verification suite runs at every deploy — deploy isn't done until it passes
  • Dedicated support, onboarding, and setup review

Enterprise

Custom annual

The entire platform self-hosted in your VPC. Nothing leaves your environment — and you can read every deployment script before you run it.

  • Full self-host: control plane, gateway, evidence, runtime
  • Deployment blueprints in plain bash — auditable, Apache-2.0
  • Audit rights and source escrow
  • Any model estate — cloud providers, private endpoints, or fully air-gapped local models
  • Signed packs with pinned trust roots
  • Support SLA and deployment assistance
  • Production Sprint included

Deploy on your terms. Every production tier ships with selective deployment blueprints — pick the one that matches your stack, and it deploys by script and verifies itself:

Single VM, full stack Your VM + hosted control plane Google Cloud Run Databricks Apps Postgres RLS data plane DynamoDB data plane Behind your existing nginx

Common questions

What happens when my balance runs out?

Requests stop — a hard stop at your balance, not an overage bill. Top up and they resume. That behavior is the product: the same enforcement that caps a runaway agent caps your account.

What is the Production Sprint?

A fixed-fee, three-week engagement. Week one, we work inside your business to find the workflow where an agent moves a real number, and baseline it. Then we deploy the agent in your environment and measure the same workflow again. Acceptance is concrete: our control checklist — per-user permissions, budget refused at the cap, kill and rollback, full action record — passes live in your environment. The fee credits toward your first year.

How is this different from an MCP gateway?

MCP gateways gate the tools; AgentHippo scopes the authority. A gateway holds a shared credential and filters which tool calls pass — behind it still sits one service account that can read everyone's rows. AgentHippo's agents never hold the credential: each turn exchanges a short-lived token for the requesting user's own authority, and your database (Postgres RLS, AWS IAM, Unity Catalog) enforces it — even if every layer above is misconfigured.

What does "governed" actually mean?

Every production deployment polls a lease from the control plane. From one console you can kill an agent, deny a user, cap spend and concurrency, or pin a signed version — effective in under a minute. Every action lands in a tamper-evident record tied to the user, session, and agent version.

All plans include the same signed IDE and CLI builds. Prices do not include applicable tax.